Search for: All records

Answering counterfactual queries has important applications such as explainability, robustness, and fairness but is challenging when the causal variables are unobserved and the observations are non-linear mixtures of these latent variables, such as pixels in images. One approach is to recover the latent Structural Causal Model (SCM), which may be infeasible in practice due to requiring strong assumptions, e.g., linearity of the causal mechanisms or perfect atomic interventions. Meanwhile, more practical ML-based approaches using naive domain translation models to generate counterfactual samples lack theoretical grounding and may construct invalid counterfactuals. In this work, we strive to strike a balance between practicality and theoretical guarantees by analyzing a specific type of causal query called domain counterfactuals, which hypothesizes what a sample would have looked like if it had been generated in a different domain (or environment). We show that recovering the latent SCM is unnecessary for estimating domain counterfactuals, thereby sidestepping some of the theoretic challenges. By assuming invertibility and sparsity of intervention, we prove domain counterfactual estimation error can be bounded by a data fit term and intervention sparsity term. Building upon our theoretical results, we develop a theoretically grounded practical algorithm that simplifies the modeling process to generative model estimation under autoregressive and shared parameter constraints that enforce intervention sparsity. Finally, we show an improvement in counterfactual estimation over baseline methods through extensive simulated and image-based experiments. 

Particle Thompson sampling (PTS) is a simple and flexible approximation of Thompson sampling for solving stochastic bandit problems. PTS circumvents the intractability of maintaining a continuous posterior distribution in Thompson sampling by replacing the continuous distribution with a discrete distribution supported at a set of weighted static particles. We analyze the dynamics of particles' weights in PTS for general stochastic bandits without assuming that the set of particles contains the unknown system parameter. It is shown that fit particles survive and unfit particles decay, with the fitness measured in KL-divergence. For Bernoulli bandit problems, all but a few fit particles decay. 

This paper proposes regenerative particle Thompson sampling (RPTS) as an improvement of particle Thompson sampling (PTS) for solving general stochastic bandit problems. PTS approximates Thompson sampling by replacing the continuous posterior distribution with a discrete distribution supported at a set of weighted static particles. PTS is flexible but may suffer from poor performance due to the tendency of the probability mass to concentrate on a small number of particles. RPTS exploits the particle weight dynamics of PTS and uses non-static particles: it deletes a particle if its probability mass gets sufficiently small and regenerates new particles in the vicinity of the surviving particles. Empirical evidence shows uniform improvement across a set of representative bandit problems without increasing the number of particles. 

A central theme in federated learning (FL) is the fact that client data distributions are often not independent and identically distributed (IID), which has strong implications on the training process. While most existing FL algorithms focus on the conventional non-IID setting of class imbalance or missing classes across clients, in practice, the distribution differences could be more complex, e.g., changes in class conditional (domain) distributions. In this paper, we consider this complex case in FL wherein each client has access to only one domain distribution. For tasks such as domain generalization, most existing learning algorithms require access to data from multiple clients (i.e., from multiple domains) during training, which is prohibitive in FL. To address this challenge, we propose a federated domain translation method that generates pseudodata for each client which could be useful for multiple downstream learning tasks. We empirically demonstrate that our translation model is more resource-efficient (in terms of both communication and computation) and easier to train in an FL setting than standard domain translation methods. Furthermore, we demonstrate that the learned translation model enables use of state-of-the-art domain generalization methods in a federated setting, which enhances accuracy and robustness to increases in the synchronization period compared to existing methodology.